I recently tried the Solaris malloc debugging facility, but the AOLserver (running ACS) went from taking about 5 minutes to start up to taking over 72 hours to start up. AOLserver uses so much dynamic memory that any malloc debugging solutions that work by adding virtual-memory hardware guard buffers around the malloc'd area are going to be r e a l l y s l o w.
I think electric fence is not thread-safe, which is a problem that a lot of the malloc debugging libraries have. Sorry I can't be more constructive. I just had a problem like this, and didn't solve it, so the system just crashes regularly.